Comment by echelon | Hacker News Reader

Comment by echelon | Hacker News Reader

Parent Full thread

Blender is amazing, and miles ahead of Gimp and other FOSS editors.

That said, I can't help but feel that all of the current generation of leading 3D software (Blender, Unreal, etc. ) is going to be replaced by something just around the corner. The progress in 3D AI is nothing short of phenomenal. It feels like soon nobody will ever have to worry about sculpting or retopologizing or rigging. An entirely new class of tool will take over.

It's not just 3D. It feels like the current generation of artistic tools (Photoshop, Illustrator, etc.) are about to be wholesale replaced with Gen AI tooling.

AI plugins (eg. the Krita plugin) are too steeped in the old world of editing. New tools will probably be AI native and prefer AI workflows for reaching the same coarse- and fine-grained outcomes.

I don't expect these tools to be used by the masses that are prompting AI slop like "50's Panavision Wes Anderson", but rather by working artists. Genuine Gen AI tools for artists.

I've been making short films on the weekends with Blender, ComfyUI, and a mix of custom software. The AI pieces are doing the heavy lifting, and my productivity is 10x what it was before Gen AI.

I think it depends on how much control you feel you want to have over your output. AI tools are decent at rendering video, but for those that want full control over the process, I feel like it's going to be a long way to go. If AI is able to generate the full source files to allow things to be modified, then I would 100% agree with you that it's going to be a new age for artists. If you're dealing with raw generated video, or even a frame by frame rendered image, your control is much more limited than if you could shape and customize your models. For most, this is probably fine, but for those that want fine-grained control, there is a far path to go in order to get to that point.

What / who is the leading edge of 3D AI? I took a look into it last year and the best tools were still a very long away from producing usable results (couldn't keep tri count under control, spiderweb rigging, shape jank, etc).

I don't care much for Generative AI (still too early) but I had some ok experiences with Tripo for retopology and auto rigging.

Still super early days though, I did have some of the same issues you had.

Thanks, I checked out Tripo. It produced some absolutely laughable trees for me, but a pretty decent looking barstool. Refused to rig it for me for some reason, though.

I'd say it'll take them some years to finish figuring out 3D, it's a good stretch beyond 2D image generation.

https://github.com/ahujasid/blender-mcp something will definitely happen we will see what happens, 3D has always been tedious so any improvements in this space will be welcome.

I totally forgot I was going to look into that...

Seems to me the 'proper' way to do something like this is to hook into makesRNA and have the endpoints automagically be generated as part of the build process. Similar (or identical) to how the python and C++ APIs are created.

I've been thinking for a while now that someone (much less lazy than myself) needs to get in there and have makesRNA generate something like IDL file(s) so any number of tools can hook into Blender's API without having to hand code the endpoints like this addon is apparently doing.

Same could be said about Fender and Gibson.

Reply on news.ycombinator.com